Inspection, aerial photography and monitoring of state and growth of protective area under power lines and other similar structures are getting increasingly popular and widespread as conventional means are either expensive and risky (using piloted aircraft) or are extremely time consuming and do not capture all the necessary angles (inspection from ground). Power line monitoring involves visual inspection of towers and their high voltage insulators, as well as the cables.
Using piloted aircraft such as expensive helicopters also require a team of flying experts using certain dedicated equipment flying close to ground level at about 10-50 knots along the physical power route. Visual inspection of linear infrastructural objects is many times limited due to certain flying regulations and other public safety concerns. In the recent past, updated techniques (post considerable improvements in flight design techniques and development of light-weight cameras) for monitoring power lines using Unmanned Aerial Vehicles (UAVs) are emerging. Benefits of using UAVs are being realized particularly with respect to costs, noise, and risk mitigation, however many drawbacks still persist as a result of limited UAV functionality. Drawbacks include over reliance on operators to observe and carry out the mission where the UAV is just used as a flying camera.
Most UAV systems lack the ability to simultaneously stream real time, encrypted observation data to multiple ground stations. Although usage of UAVs for maintenance inspections, especially of long linear infrastructure is rapidly emerging as a popular option, the amount of video or image data acquired is typically huge, due to vastness of infrastructure. Thus, automated analysis of such images and videos are being increasingly sought. Such analysis necessitates detection of elongated foreground objects, commonly subjected as linear feature detection.
Conventional techniques so far for implementing automated linear feature detection in outdoor scenes have used Hough transform for clustering of lines. The Hough transform is a technique which can be used to isolate features of a particular shape within an image. Since Hough transform requires that the desired features be specified in some parametric form, the classical Hough transform is most commonly used for the detection of regular curves such as lines, circles, ellipses, etc. A generalized Hough transform can be employed in applications where a simple analytic description of a feature(s) is not possible.
However, the Hough transform has a computational complexity of the order of O(n3) which is considerably high. Additionally, the power lines range for long distances over different terrains vary and hence the background imagery can vary from trees and patches of greenery to different flat spaces along with other common objects such as homes and roads. Detection of long linear infrastructural objects is challenging but necessary.
A couple of conventional techniques use the process of canny edge detection as the core step and implementation of canny edge detection leads to higher computational complexity and this higher computational complexity is not suitable for near-real-time processing. Other prior techniques either tackle their problem of linear feature detection in medical imaging domain (where the background imagery is plain and simple and are not applicable for outdoor scene analysis directly) or are not robust enough to detect false positives.
In general, existing techniques have assumed the background imagery to be near-stochastic, and accordingly focused primarily at background modeling and subtraction. This generalist assumption simplifies the process design to a rudimentary level as in a realistic scenario, outdoor video background is not stochastic (i.e having a random probability distribution or pattern that may be analyzed statistically but may not be predicted precisely) and hence multiple ways of background suppression rather than (complete) subtraction needs to be evolved.